> Bear in mind that current sites rely greatly on static content such as
> css files, JavaScript files and (planty of) images. I find it hard to
> believe that Tomcat or Resin will do better job than Apache.
Nobody claimed that. The claim was in fact the other way round: Apache is
better at serving static content. But you need Tomcat / Resin for the
dynamic part if that is to be done in JSP / Servlet technology.
> Also, don't
> you want to save the JVM's CPU and memory fo dymanic content rather than
> static one?
Err, the CPU is the same as that of the Apache. :-) And memory overhead
is not that much for static content (it's streamed). All in all IMHO
Henrik is completely right to suggest to use the tandem only for heavy
load sites. Tomcat / Resin are not that bad at serving static content;
after all, that is mostly IO bound and that depends on the OS, disks and
network connection. Same for Apache and Tomcat.
> I's suggest to do a small benchmark - see what is the porformance
> difference between the options (It would be also nice to hear about your
> impressions of the two settings).
It's always a good idea to do benchmarking to have concrete numbers for
comparison. Note however that it is not too easy to simulate a high
traffic web site. The benchmark itself can get complicated and expensive
soon.
Regards
robert
> David
>
[quoted text clipped - 16 lines]
> >>Coaching, problem solving, economical contract programming.
> >>See http://mindprod.com/jgloss/jgloss.html for The Java Glossary.
Mohun Biswas - 29 Mar 2004 15:18 GMT
> "David Rabinowitz" <dar@cs.huji.ac.il> schrieb im Newsbeitrag
>>Also, don't
>>you want to save the JVM's CPU and memory fo dymanic content rather than
>>static one?
>
> Err, the CPU is the same as that of the Apache. :-)
Actually there's no requirement that the two servers run on the same
machine, let alone the same CPU.
MB
Robert Klemme - 29 Mar 2004 15:53 GMT
> > "David Rabinowitz" <dar@cs.huji.ac.il> schrieb im Newsbeitrag
> >>Also, don't
[quoted text clipped - 5 lines]
> Actually there's no requirement that the two servers run on the same
> machine, let alone the same CPU.
Oops, yeah that's true. Thanks for correcting that!
robert
David Rabinowitz - 29 Mar 2004 16:51 GMT
>>Bear in mind that current sites rely greatly on static content such as
>>css files, JavaScript files and (planty of) images. I find it hard to
[quoted text clipped - 3 lines]
> better at serving static content. But you need Tomcat / Resin for the
> dynamic part if that is to be done in JSP / Servlet technology.
As I see it, the claim was in the opinion that can drop Apache from the
system.
>>Also, don't
>>you want to save the JVM's CPU and memory fo dymanic content rather than
[quoted text clipped - 6 lines]
> after all, that is mostly IO bound and that depends on the OS, disks and
> network connection. Same for Apache and Tomcat.
Not exactly - Apache, written in C, will be faster than Java serving
static content. Also, why waste a tomcat working thread (forgot to
mention it earlier...) on static content?
>>I's suggest to do a small benchmark - see what is the porformance
>>difference between the options (It would be also nice to hear about your
[quoted text clipped - 4 lines]
> traffic web site. The benchmark itself can get complicated and expensive
> soon.
So I guess we are all waiting anxiously ;-) Does anybody wants to peek
the glove?
David
Mohun Biswas - 29 Mar 2004 17:37 GMT
> As I see it, the claim was in the opinion that can drop Apache from the
> system.
The record is quite clear. He said "For an all-out dynamic site you
might just as well run tomcat or resin standalone". Meaning that Apache
cannot help you with dynamic java-based resources. There's nothing to
argue with here; it's just plain true.
MB
Mohun Biswas - 29 Mar 2004 17:49 GMT
>> As I see it, the claim was in the opinion that can drop Apache from
>> the system.
[quoted text clipped - 3 lines]
> cannot help you with dynamic java-based resources. There's nothing to
> argue with here; it's just plain true.
Also, from the Resin website
(http://www.caucho.com/resin-3.0/install/httpd.xtp):
> Resin provides a fast standalone web server. In our benchmarks, it serves static pages faster than Apache!
(I have not independently evaluated this claim, nor ever used resin for
that matter).
MB
Jon A. Cruz - 29 Mar 2004 18:46 GMT
> Not exactly - Apache, written in C, will be faster than Java serving
> static content. Also, why waste a tomcat working thread (forgot to
> mention it earlier...) on static content?
If you're basing the speed issue on having been written in C, that's not
really a good measurement. More depends on the networking and *how* the
app was written, not which language it was written in.
Robert Klemme - 30 Mar 2004 13:58 GMT
> >>Bear in mind that current sites rely greatly on static content such as
> >>css files, JavaScript files and (planty of) images. I find it hard to
[quoted text clipped - 6 lines]
> As I see it, the claim was in the opinion that can drop Apache from the
> system.
... if and only if the amount of static content served is considerable
small compared to the dynamic content. That was the claim. I added that
the crucial factor in serving static content is disk speed, the filesystem
and network bandwidth - not so much the language or technology used to
implement the server.
> Not exactly - Apache, written in C, will be faster than Java serving
> static content.
Not necessarily (see above).
> Also, why waste a tomcat working thread (forgot to
> mention it earlier...) on static content?
Why not? A tomcat thread is cheaper than a complete Apache process.
robert